New class of time-periodic solutions to the 1D cubic wave equation

Filip Ficek; Maciej Maliborski

REVIEW 3 major objections 6 minor 2 cited by

New class of time-periodic solutions to the 1D cubic wave equation

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper rigorously constructs three distinct time-periodic solutions of the defocusing cubic wave equation on an interval, two of them belonging to a new 'branch' class beyond the classical trunk family.

desk verdict A clean analytic framework with a tight, code-delivered computational certificate; the existence proof is believable but the margins are thin enough that the scripts deserve a careful referee. read the letter →

arxiv 2506.10839 v1 pith:5HYO55NL submitted 2025-06-12 math.AP math-phmath.MP

Filip Ficek , Maciej Maliborski This is my paper · ORCID

classification math.APmath-phmath.MP MSC 35B1068V0535B3235L71

keywords Time-periodicsolutionsNonlinearwaveequationBifurcationsComputer-assistedproofRationalarithmeticFixed-pointargumentCubicDirichletboundaryconditions

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proves that the one-dimensional defocusing cubic wave equation $u_{tt}-u_{xx}+u^3=0$ with Dirichlet boundary conditions has at least three pairwise distinct $2\pi$-periodic solutions, all with the same frequency $\Omega=69/40$ after time rescaling. One of these sits on the classical 'trunk' family of solutions known since the 1980s, while the other two have larger energy and stronger high-mode content, and belong to a conjectured 'branch' structure. The proof is a computer-assisted fixed-point argument: it starts from approximate solutions $u_0^{(i)}$ with rational Fourier coefficients, constructs a finite-dimensional correction operator $A$, and verifies with exact rational arithmetic that a contraction mapping on a weighted $\ell^1$ space has a fixed point close to each $u_0^{(i)}$. If correct, this is the first rigorous confirmation that the solution set of this equation is richer than the classical Cantor-like family.

What carries the argument

The argument works in the Banach space $X$ of $2\pi$-periodic functions spanned by $P_{m,n}(\tau,x)=\cos((2m+1)\tau)\sin((2n+1)x)$, with norm $\|v\|=\sum \rho_\tau^{2m+1}\rho_x^{2n+1}|\hat v_{m,n}|$ and weights $\rho_\tau=\rho_x=1+10^{-20}$. The load-bearing object is the linear part $H_0(h)=-3L_\Omega^{-1}(u_0^2 A h)+h-Ah$ of the fixed-point map $N_\Omega(h)=F_\Omega(u_0+Ah)-u_0+(I-A)h$, where $F_\Omega(u)=-L_\Omega^{-1}u^3$ and $A$ is a finite-dimensional rational matrix approximating the inverse of $I+3L_\Omega^{-1}\Lambda_{u_0^2}$. Lemma 3 gives the explicit bound $\|L_\Omega^{-1}v\|\le \phi(m,n)\|v\|$ with $\phi(m,n)=4q^2/(2\max(2q(2n+1),(2p+1)(2m+1))-1)$, and the choice $\Omega=(2p+1)/(2q)$ makes the denominators differences of an even and an odd integer, so the small-divisor problem---the near-resonances that plague generic perturbative constructions---does not arise. Lemma 4 bounds the action of $L_\Omega^{-1}(u_0^2P_{m,n})$ on high modes, and formula (5.2) reduces $\|H_0\|$ to a finite maximum that the supplied scripts evaluate exactly in rational arithmetic.

What would settle it

Rerun the supplied scripts with the stated data (or independently recompute the maxima in formula (5.2) using interval arithmetic) and check whether, for each $i$, the quantity $\|H_0\|+6\|L_\Omega^{-1}\|\|u_0^{(i)}\|\|A\|^2\delta+3\|L_\Omega^{-1}\|\|A\|^3\delta^2$ is smaller than $K_0$ and $\|N_\Omega(0)\|$ is smaller than $(1-K_0)\delta$. If any of these inequalities fails, the contraction argument does not close and the claimed solutions are not established.

Watch

Extended reading notes

Core claim

For frequency $\Omega=69/40$, the paper's Theorem 1 asserts that for each of three explicitly given rational-coefficient approximate solutions $u_0^{(i)}$, $i=1,2,3$, there is an exact solution $u^{(i)}$ of the rescaled equation satisfying $\|u^{(i)}-u_0^{(i)}\|<\varepsilon^{(i)}$ in a weighted $\ell^1$ norm, with $\varepsilon^{(1)}\approx 1.79\times 10^{-8}$, $\varepsilon^{(2)}\approx 1.40\times 10^{-8}$, and $\varepsilon^{(3)}\approx 2.18\times 10^{-7}$; the six functions $\pm u^{(i)}$ are pairwise distinct. One solution is dominated by the lowest mode and lies on the known trunk family, whereas the other two carry substantial higher-mode content and are the first rigorously constructed members of the new branch class conjectured from numerical Galerkin computations. A symmetry argument then yields time-periodic solutions of the focusing equation $u_{tt}-u_{xx}-u^3=0$ with frequency $40/96$.

Load-bearing premise

The proof rests on the correctness of the supplied computer scripts, which check the inequalities (2.6) by exact rational arithmetic; if a script or the formula (B.6)/(B.7) for the norm of $H_0$ contains an implementation error, the bounds and hence the existence conclusions could fail.

Editorial extensions

If this is right

There exist at least three pairwise distinct $2\pi$-periodic solutions of the rescaled equation (1.2) at frequency $\Omega=69/40$, and hence of the original equation (1.1) after undoing $\tau=\Omega t$.
Two of these solutions are the first rigorously confirmed members of the branch family that had previously been seen only numerically.
The same symmetry gives time-periodic solutions of the focusing cubic wave equation with frequency $40/96$.
The rational-frequency condition $\Omega=(2p+1)/(2q)$ makes the denominators in $L_\Omega^{-1}$ odd, so the same finite rational verification scheme is not blocked by small divisors at any such frequency.
The explicit bounds $\varepsilon^{(i)}$ locate actual solutions inside tiny balls around rational approximations, giving quantitative control of the construction.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Editorial extension: if the branch pairs persist at other rational frequencies of the form (1.3), the solution set would contain infinitely many such pairs accumulating along the trunk, matching the fractal-like picture suggested by numerics.
Editorial extension: the same exact-rational-arithmetic verification scheme should transfer to other 1D semilinear wave equations with polynomial nonlinearities, since it only needs product-to-sum identities and an explicit $\phi(m,n)$ bound for the linear resolvent.
Editorial extension: a continuation in $\Omega$ from $69/40$ could test whether the two branch solutions remain close to their numerical approximations, and whether new branch pairs appear at nearby rational frequencies.
Editorial extension: the tiny radii $\varepsilon^{(i)}$ suggest the approximate Galerkin solutions are very accurate; a reader could rerun the supplied scripts at higher truncation to look for additional branch solutions at the same frequency.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The new thing here is that two solutions on the conjectured branch-family of the 1D cubic wave equation are now proven to exist, not just numerically indicated. The analytic setup is genuinely solid: the weighted l1 space, the contraction argument, and the lemmas bounding the nonlinearity and the inverse of the linear operator are all self-contained and correct. Lemma 1 (the product bound) and Lemma 3 (the rational-frequency inverse bound) are particularly clean. The paper also does something rare: it is honest about what is and is not proved. Remarks 1 and 2 put the single-frequency result in context and do not overclaim. The real load-bearing part is the computer verification of Theorem 2. The supplied Mathematica scripts and data files make the exact rational arithmetic reproducible in principle, and the formulas in Appendix B are detailed enough to reimplement. That is exactly what a computer-assisted proof should provide. Where I would push back is the tightness of the margins. For the two branch solutions, K0 - ||H0|| is about 5e-5. That is not a lot of room. If any script contains an error - especially in the g-coefficients from (B.4) or the norm evaluation in (B.6)-(B.7) - the printed rational bounds could be wrong and the inequality (2.6) could fail. The paper would be stronger if the certificate were frozen (checksums or a commit hash) and if the scripts explicitly verified that A is invertible, as Theorem 2 assumes. I suspect invertibility is not actually needed for the fixed-point argument to produce a solution, since the fixed point of N still gives u = u0 + Ah, but the paper asserts it and should check it. I have not rerun the scripts. If I were the referee, I would ask the authors to confirm the run environment, provide a frozen version of the code, and ideally get an independent reimplementation of the norm computation. That said, I see no red flag that suggests the computation is wrong; the risk is the usual one for computer-assisted proofs, and it is properly disclosed. This paper deserves a serious referee. It is an honest, technically sound advance in a small but active subfield. The contribution is modest but real, and the methods are worth having on the record. Send it to review, but make sure the referee actually engages with the computer-assisted part.

Referee Report

3 major / 6 minor

Summary. The paper proves the existence of three nontrivial 2π-periodic solutions to the defocusing cubic wave equation on an interval with Dirichlet boundary conditions, all at frequency Ω = 69/40. The proof combines a weighted ℓ1 fixed-point framework with a computer-assisted verification of operator bounds using exact rational arithmetic. One solution is close to a low-mode Galerkin approximation; the other two are close to higher-energy approximations, supporting the authors' earlier numerical conjectures about 'branches' beyond the classical 'trunk' family. The analytic part is self-contained and reduces the existence claim to Theorem 2, whose hypotheses are verified by supplied Mathematica scripts and data files.

Significance. If the computational certificate is correct, the result is a valuable contribution to the rigorous theory of time-periodic solutions of resonant 1D nonlinear wave equations, providing the first proof of solutions outside the classical Cantor-family trunk. The analytic framework is clean: Lemmas 1–4 give transparent bounds on multiplication, the inverse of L_Ω, and tail estimates, and the reduction to the finite verification (2.6) is rigorous. The use of exact rational arithmetic is a sound choice for computer-assisted proofs and avoids rounding-error concerns. The authors are to be commended for supplying the code and data. The main risk is that the existence claim rests entirely on the correctness of unverified scripts with very tight margins.

major comments (3)

[Section 7 and Appendix A] The proof of Theorem 2 is delegated to the Mathematica scripts, but the manuscript does not include the output logs or an explicit evaluation of the inequalities (2.6). The margins are extremely tight: for the first and second solutions, K0 − ||H0|| is approximately 5.2×10^-5 and 4.1×10^-5, respectively, which is of the same order as the δ-dependent terms in (2.6). A small error in the scripts' evaluation of (B.6)–(B.7) or in the g-coefficients from (B.4) could flip the inequality and invalidate Theorem 1. Please provide a machine-readable certificate of all rational bounds and the final inequalities, including exact values of ||N(0)|| and of the left-hand side of the first inequality in (2.6), and freeze the exact code and data versions (e.g., with checksums).
[Section 4] Theorem 2 assumes that A is a linear isomorphism, but the construction of A as a rationalized approximate floating-point inverse of Ã does not include a proof that the resulting matrix A (and hence the block-diagonal operator A) is invertible. The text says the matrix is 'sufficiently close' to an inverse, but no exact determinant or explicit inverse is provided. Please supply an exact rational determinant or an explicit inverse for each A^(i), or alternatively modify Theorem 2 and the proof of Theorem 1 to remove the invertibility hypothesis, since the contraction argument itself only requires A to be a bounded linear operator.
[Appendix B] The convolution formulas (B.2)–(B.4) are central to assembling the matrix A and computing the bound on ||H0||, yet they are stated without derivation. An off-by-one error or a sign error in these formulas would propagate directly into the claimed bounds. Please add a derivation or a reference to one, and include a validation script that checks (B.2)–(B.4) against direct symbolic trigonometric products for small random inputs, so that the correctness of these formulas is independently verifiable.

minor comments (6)

[Section 2.3] The contraction argument is applied on the open ball Bδ(0), which is not a complete metric space. Since the inequalities in (2.6) are strict, the map sends the closed ball into itself; the proof should be formulated on the closed ball to apply the Banach contraction principle.
[Lemma 1 proof] In the final line of the proof, the last sum is written with |\hat u_{m1,n1}|; it should be |\hat w_{m3,n3}| to match the product of the three norms.
[Lemma 2 proof] In the display after the first inequality, the supremum is written with P_{m1,n1} and the denominator ρ^{2m+1}ρ^{2n+1}; the subscripts m1,n1 should be m,n for consistency.
[Section 7] The ε^(i) values are given only as decimal expansions in (7.1). Since the paper emphasizes exact rational arithmetic, please provide rational upper bounds for these quantities or state explicitly that the decimals are rigorous upper bounds computed from the rational data.
[Appendix A] Please specify the exact version of Wolfram Mathematica used, the operating system, and the hardware environment, and provide checksums for the data files and scripts to allow exact reproduction.
[Lemma 3] The formula for φ(m,n) is typeset ambiguously in the text due to line breaking; write it as a single fraction, \frac{4q^2}{2\max(2q(2n+1),(2p+1)(2m+1))-1}, to avoid confusion.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: prior numerical papers motivate candidate solutions, but the existence proof is a self-contained contraction argument with explicit rational bounds.

full rationale

The paper's central claim is Theorem 1, and it is derived from Theorem 2 via a standard Banach fixed-point argument. The approximate solutions u0^(i) and the frequency Ω are taken from the authors' earlier numerical work [FM25; FM24], but those works are used only to locate candidate solutions; the existence proof does not assume any existence conclusion or any property of the true solution. All input data are explicit rational numbers in the Supplemental Material, and the proof verifies the operator norm bounds and inequalities (2.6) using exact rational arithmetic. The operator A is explicitly constructed as an approximate inverse of I + 3L_Ω^{-1}Λ_{u0^2}; this is a legitimate proof device, not a concealment of the conclusion. Choosing K0 and δ after computing the bounds is standard witness construction in computer-assisted proofs, not circular fitting, because the inequalities are verified after substitution. Self-citations to [FM25] and [FM24] are motivational and heuristic; none of the load-bearing inequalities or fixed-point hypotheses rely on the truth of those papers. The computational certificate is not frozen and the scripts are not independently machine-checked, but that is a reproducibility and correctness risk, not a circularity of the mathematical argument. No equation is defined in terms of the target result, no fitted parameter is renamed as a prediction, and no uniqueness or ansatz is imported via self-citation. Therefore no significant circularity is present.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The proof imports from prior theory only standard functional analysis and the PDE setup; its concrete inputs are six proof constants plus three rationalized numerical approximations, all chosen to close the contraction inequalities. No new physical entity is introduced.

free parameters (6)

Norm weights ρτ and ρx = 1+10^-20
Section 6 states no attempt was made to optimize; every operator norm bound depends on this choice.
Galerkin truncation M=N for each u0^(i) = 36, 38, 13 for i=1,2,3
Appendix A.3 lists runtimes; selected so that ||NΩ(0)|| is small enough for (2.6).
Matrix block size μ=ν for A = not stated in text, supplied in data files
Section 4 chooses μ=ν minimally so that the first term in (5.2) is below 1.
Tail truncations M̃ and Ñ = not stated in text, supplied in data files
Appendix B.3 sets them as the smallest values making the last two terms in (5.2) less than 1.
Contraction constants K0 and δ = rational fractions in Section 7
Proof parameters, not physical constants; chosen after all bounds to satisfy (2.6).
Rational approximate solution coefficients u0^(i) = rational vectors in Supplemental Material
Remark 1: coefficients of numerical solutions from [FM25] were replaced by close rational numbers; Theorem 1 asserts existence near these exact vectors.

assumptions (5)

standard math X is a Banach space and satisfies the algebra-type bound ||uvw|| ≤ ||u|| ||v|| ||w|| (Lemma 1).
Proved in Section 3 from the weighted ℓ1 norm; needed for treating the cubic term as a bounded map.
standard math LΩ has a bounded inverse on X with the stated φ(m,n) bounds (Lemma 3).
Used in the fixed-point formulation; relies on the rational form of Ω making all denominators odd and hence nonzero.
standard math The Banach contraction principle applies to NΩ on the closed ball Bδ(0).
Gives the fixed point h and the distance estimate ||u-u0|| ≤ ||A||δ in Section 2.3.
domain assumption The rational arithmetic implementation in Mathematica correctly computes the formulas in Appendix B, including (5.2), (B.6), and (B.7).
Load-bearing for Theorem 2; no formal proof or commit hash is supplied for the software.
domain assumption Restriction to frequencies Ω=(2p+1)/(2q) with p>q, so Ω>1 and small divisors are avoided.
The proof is carried out only for Ω=69/40; this arithmetic structure is used in Lemma 3 and in the absence of small divisors.

how reviews work

0 comments

Cite this review

Pith. "Pith review of New class of time-periodic solutions to the 1D cubic wave equation." pith.science (2026). https://pith.science/paper/5HYO55NL

@misc{pith2026250610839,
  author       = {Pith},
  title        = {Pith review of: New class of time-periodic solutions to the 1D cubic wave equation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5HYO55NL}},
  note         = {Machine review of arXiv:2506.10839}
}

read the original abstract

In recent papers (arXiv:2407.16507, arXiv:2408.05158) we presented results suggesting the existence of a new class of time-periodic solutions to the defocusing cubic wave equation on a one-dimensional interval with Dirichlet boundary conditions. Here we confirm these findings by rigorously constructing solutions from this class. The proof uses rational arithmetic computations to verify essential operator bounds.

Figures

Figures reproduced from arXiv: 2506.10839 by the authors.

**Figure 1.** (Left) Frequency diagram of time-periodic solutions computed with a mode truncation N = M = 9 (cf. [FM24]), illustrating the global bifurcation structure. (Right top) A zoomed-in region around Ω = 69/40 = 1.725, showing the locations of the three approximate solutions u (i) 0 , i = 1, 2, 3, studied in this work. (Right bottom) The same region plotted with respect to the weighted ℓ 1 norm instead of energy E, see Eq… view at source ↗

**Figure 2.** Diagram presenting spaces YM,N , XM,0, and X0,N for some fixed M and N. Dots represent basis functions Pm,n belonging to appropriate subspaces. The coordinate axes intersect at the point (0, 0). We will denote an open ball of centre v ∈ X and radius r > 0 with respect to the norm ∥ · ∥ as Br(v). For two non-negative integers M and N, we define the following subspaces of X: XM,N := {v ∈ X : ˆvm,n = 0 for m < M or n … view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Nonlinear oscillations of strings and beams
math-ph 2025-08 conditional novelty 6.0 of 10

Fractal-like families of stable, large-energy multi-mode periodic solutions are found in the cubic wave and beam equations via Galerkin continuation and reducible mode analysis.
Time-periodic solutions of the conformally invariant wave equation on the Einstein cylinder
gr-qc 2025-08 conditional novelty 3.0 of 10

For the cubic conformal wave equation on the Einstein cylinder, time-periodic solutions form complex trunk-branch bifurcation networks, and a resummed Poincaré-Lindstedt series is claimed to encode these structures.

Reference graph

Works this paper leans on

3 extracted references · 2 canonical work pages · cited by 2 Pith papers

[1]

Arioli and H

[AK17] G. Arioli and H. Koch. Families of Periodic Solutions for Some Hamiltonian PDEs.SIAM Journal on Applied Dynamical Systems, 16(1):1–15, 2017.doi: 10.1137/16m1070177. [AKN88] V. Arnold, V. Kozlov, and A. Neishtadt.Dynamical systems III. V. I. Arnold, editor. Encyclopaedia of Mathematical Sciences. Springer Berlin, Heidelberg, 1988.doi:10.1007/978-3-6...

doi:10.1137/16m1070177 2017
[2004]

[GMP05] G

doi:10.1016/j.matpur.2004.01.007. [GMP05] G. Gentile, V. Mastropietro, and M. Procesi. Periodic Solutions for Completely Resonant Nonlinear Wave Equations with Dirichlet Boundary Conditions.Com- munications in Mathematical Physics, 256:437–490, 2005.doi:10.1007/s00220- 004-1255-8. [LS88] B. V. Lidskii and E. I. Shul’man. Periodic solutions of the equation...

work page doi:10.1016/j.matpur.2004.01.007 2004
[2008]

REFERENCES 18 [DS24] R

doi:10.1016/j.aim.2007.11.004. REFERENCES 18 [DS24] R. Donninger and B. Sch¨ orkhuber. Self-similar blowup for the cubic Schr¨ oodinger equation.arXiv, 2024.doi:10.48550/arXiv.2406.16597. [FM24] F. Ficek and M. Maliborski. Trees, trunks, and branches - bifurcation structure of time-periodic solutions tou tt −u xx ±u 3 = 0.arXiv, 2024.doi:10.48550/ arxiv.2...

work page doi:10.48550/arxiv.2406.16597 2007

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Arioli and H

[AK17] G. Arioli and H. Koch. Families of Periodic Solutions for Some Hamiltonian PDEs.SIAM Journal on Applied Dynamical Systems, 16(1):1–15, 2017.doi: 10.1137/16m1070177. [AKN88] V. Arnold, V. Kozlov, and A. Neishtadt.Dynamical systems III. V. I. Arnold, editor. Encyclopaedia of Mathematical Sciences. Springer Berlin, Heidelberg, 1988.doi:10.1007/978-3-6...

doi:10.1137/16m1070177 2017

[2] [2004]

[GMP05] G

doi:10.1016/j.matpur.2004.01.007. [GMP05] G. Gentile, V. Mastropietro, and M. Procesi. Periodic Solutions for Completely Resonant Nonlinear Wave Equations with Dirichlet Boundary Conditions.Com- munications in Mathematical Physics, 256:437–490, 2005.doi:10.1007/s00220- 004-1255-8. [LS88] B. V. Lidskii and E. I. Shul’man. Periodic solutions of the equation...

work page doi:10.1016/j.matpur.2004.01.007 2004

[3] [2008]

REFERENCES 18 [DS24] R

doi:10.1016/j.aim.2007.11.004. REFERENCES 18 [DS24] R. Donninger and B. Sch¨ orkhuber. Self-similar blowup for the cubic Schr¨ oodinger equation.arXiv, 2024.doi:10.48550/arXiv.2406.16597. [FM24] F. Ficek and M. Maliborski. Trees, trunks, and branches - bifurcation structure of time-periodic solutions tou tt −u xx ±u 3 = 0.arXiv, 2024.doi:10.48550/ arxiv.2...

work page doi:10.48550/arxiv.2406.16597 2007